What the 2026 AI Index Reveals About Governance Gaps and Global Policy Shifts

Responsible AI Lags Behind Rapid Capability Gains

The report finds that while almost every leading AI lab publishes comprehensive capability benchmark scores, few disclose safety, fairness, or privacy metrics, indicating a clear lag in responsible‑AI reporting.

AI Incidents Rise Sharply

Recorded AI‑related incidents worldwide jumped from 233 in 2024 to 362 in 2025, a rise of over 50%. The data also shows that improving one responsibility dimension (e.g., safety) can sometimes degrade another (e.g., accuracy), making remediation complex.

Policy Divergence Across Regions

In 2025 the EU’s AI Act began enforcing its first set of prohibitions, while the United States moved toward regulatory relaxation. Japan, South Korea, and Italy each passed national AI laws. Notably, more than half of the newly announced AI strategies come from developing countries, marking their first large‑scale policy entry.

AI Sovereignty Becomes a Global Consensus

Countries are increasingly focused on securing their own AI ecosystems to avoid dominance by a few major powers, investing in super‑computing centers and domestic regulatory frameworks.

Talent Flow Shows a Steep Decline in US Attractiveness

Although the US remains the largest hub for AI talent, the number of researchers and developers moving to the US has fallen 89% since 2017, with an 80% drop in the most recent year, representing one of the sharpest declines in recent history.

Simultaneously, open‑source contributions are shifting: non‑Western regions now surpass Europe on GitHub and are closing the gap with the US, leading to more diverse language models and benchmarks.

Expert‑Public Perception Gap

When asked about AI’s impact on jobs, 73% of AI experts view it positively, compared with only 23% of the general public—a 50‑point difference. Similar gaps appear in economic and healthcare domains.

Trust in government AI regulation is lowest in the United States at 31%, while the EU enjoys the highest global trust levels, highlighting a stark contrast in public confidence.

Conclusion: Balancing Speed and Governance

The data does not point to a single trajectory: AI technology accelerates while governance structures scramble to adapt. Optimism and anxiety coexist, and the key challenge is finding a balanced approach that ensures AI serves humanity without outrunning oversight.